Apple to figure out your typing speed to improve autocorrect

27comments
Apple to figure out your typing speed to improve autocorrect
What can you say about Apple's autocorrect? Sometimes too zealous to the point of being obscene, it still is considered one of the best in the business. Now, a trio of patent applications reveals some ideas Apple has for improving your typing on the virtual iPhone keyboard. Patent Application No. 12/976834 has to do with "Combining timing and geometry information for typing correction." The idea is to use keystroke geometry and timing to discern the typist's "intended input".

This patent uses a timestamp on a string of typed characters which is then analyzed using a process in autocorrect or autocompletion. A "baseline typing speed" can be computed and if it takes longer to type a particular sentence, the delay could mean that there are misspelled words or other typing errors. Speed faster than the baseline could indicate the double strike of a letter by mistake. The application reads, "For example, if the string “theere” is entered and the time between the keystrokes “ee” is less than an associated baseline by more than a threshold amount, the replacement candidate “there” may be assigned a higher score."

The whole technology could be put into the Apple iPhone and start with a pre-determined typing speed that would be adjusted for the phone's actual user and if the user typed faster or slower than the typical user. Besides using speed to catch mistakes, the feature would also examine the location of the letters typed on the keyboard and the geometry of words spelled. words. The patent application reads, "For example, a word that is similar to the typed text except for one or more errors associated potentially with keyboard geometry, such as differing by a letter where the correct letter is located adjacent to the typed incorrect letter on the keyboard, may be suggested."

Recommended Stories
A second autocorrect-related patent application, No. 12/976849, is entitled "Using parts-of-speech tagging and named entity recognition for spelling correction." This method involves understanding the context of typed words so as to understand the context of certain tagged words. For example, knowing the context of  a sentence can help autocorrect know if the use of the word Apple in a sentence refers to Apple the tech company, or apple the delicious fruit. This patent requires the use of a complex statistical language model and the process involves parsing words, phrases and sentences while looking for capitalization, prefixes and suffixes.

The third patent, No. 12/976864, also deals with trying to "tag" certain parts of written speech but limits the language processing to search queries. The patent application discusses a process for analyzing phrases of speech to make searching documents or the internet, more efficient. Some of Apple's autocorrect patents have been used in its lawsuits with Samsung. Back in February, Apple accused Samsung of infringing on a patent for a "Method, system, and graphical user interface for providing word recommendations".

source: AppleInsider
 


Recommended Stories

Loading Comments...
FCC OKs Cingular\'s purchase of AT&T Wireless